[ESSI-2062] fix iiif urls unaware of external_storage

[aploshay]

Refactors lookup of IIIF image server URLs (in a manner aware of external_storage settings), and applies that logic where not yet present.

New/modified infrastructure:

• app/services/iiif_file_set_path_service.rb (new, replaces direct calls to Hyrax.config.iiif_image_url_builder, Hyrax.config.iiif_info_url_builder)
• app/models/file_set.rb (modified)
• Hyrax::FileSetPresenter modified via an extension

Updated cases, adding handling of external storage:

• app/models/collection_branding_info.rb (part of initial bug report generating ESSI-2062)
• app/services/essi/generate_pdf_service.rb (not tested, but expected to be broken for external storage files)
• app/services/iiif_collection_thumbnail_path_service.rb (part of initial bug report generating ESSI-2062)

Refactoring cases:

• app/controllers/purl_controller.rb
• app/helpers/catalog_helper.rb
• app/presenters/hyrax/displays_image.rb
• app/services/iiif_thumbnail_path_service.rb

[dlpierce]

Pulling these IIIF behaviors into this service presents an opportunity to make some spec tests.

[aploshay]

@dlpierce spec tests added, which did indeed lead to some revisions!

[aploshay]

@dlpierce for the refactoring pass, I didn't want to change any logic, but there's an opportunity to simplify/unify logic if it was unintentionally variable, originally:

• if we get consistent (either way) about just checking that content_location has text (at all) vs starting with "s3", we can simplify #versioned_lookup_id to call #basic_lookup_id
• if we get consistent about checking for file versioning if #basic_lookup_id fails -- originally done only in the DisplaysImage module -- we can do away with the #basic_lookup_id vs #versioned_lookup_id split, entirely

[aploshay]

@dlpierce so that for failing test, it's complaining that it's not expecting version information in the indexed path:
...%2Ffcr:versions%2Fversion1/full/250,/0/default.jpg
vs
.../full/250,/0/default.jpg (expected)

What are your thoughts how to approach a stable fix?

• we could amend the test, so the matching pattern passes with or without version information
• we could amend the test, to expect version information (and fail without)
• we could amend FileSet#original_file_id so it doesn't do any version lookup -- the test should pass as-written, and that might be more in-line with expectations?

[dlpierce]

Looking good! Thanks for cleaning this up!